Framework Laravel Package

Product Decisions This Supports

• Build vs. buy for AI-assisted development on a proprietary stack. Tchiowa (salaab/framework) is a Laravel-shaped framework without laravel/framework, which means the official Laravel Boost and its ecosystem guidance do not apply. v1.1.0 answers the "do we hand-roll agent context or adopt a maintained bundle?" question in favor of adoption: 7 guidelines (core, php-strict, illuminate, bladeone, phinx, frontend, pest) and 3 skills (crud-tchiowa, frontend-tchiowa, phinx-migration) are now shipped rather than tribal knowledge.
• Onboarding and time-to-first-PR. The guidelines + skills set is effectively executable documentation. This is a lever on ramp-up cost for new hires and contractors, and on the quality floor for AI-generated code (fewer hallucinated Eloquent/Artisan idioms in a codebase that has neither).
• Agent-to-runtime integration as a roadmap bet. The Node stdio MCP server (mcp/tchiowa/, src/index.ts + built dist/, 4 tools) registered across .mcp.json, opencode.json, and tchiowa.json turns the coding agent from a text generator into something that can inspect the project. This is the differentiating feature of the release and the one worth tracking adoption on.
• Standardization as a product decision. pint.json (Laravel preset) and phpstan.neon at level 5 set an enforceable quality bar. Committing to this package means committing to those defaults or explicitly diverging — decide which, don't drift.
• Operational safety surface. The new DatabaseExporter.php plus 17 console commands give agents and humans a sanctioned path to inspect schema/data instead of ad-hoc DB access. Treat this as a governance/security feature, not just a convenience.
• Risk item to schedule, not discover. BladeFactory now defaults to MODE_AUTO and helpers gained guards — behavior-changing defaults inside a "guidelines and tooling" release. Budget a regression pass on view resolution and helper call sites before rollout.

When to Consider This Package

Adopt when all of the following hold:

• You are on the Tchiowa fork (salaab/framework) with the documented stack: PHP ^8.4, illuminate/* ^10.x, BladeOne ^4.9, Phinx ^0.13, Vite, Tailwind. The package is purpose-built for exactly this matrix and is not portable.
• Your team uses AI coding agents (Claude Code, opencode, or any MCP-capable client) on this codebase, and you want repo-specific context rather than generic Laravel advice.
• You are willing to run a Node sidecar alongside PHP for the MCP server and to maintain the .mcp.json / opencode.json / tchiowa.json registrations as part of your dev environment.
• You want a shared baseline for formatting and static analysis and are prepared to enforce it in CI.
• You accept single-maintainer, community-scale support and can vendor/pin the package (note the release is a squashed merge of PR #2 — audit the diff before pinning).

Look elsewhere when:

• You are on stock Laravel — use official Laravel Boost; this package's value (Laravel idioms adapted to a non-Laravel runtime) inverts into misfit.
• Your stack uses Eloquent migrations rather than Phinx, or Inertia/Livewire rather than BladeOne — the skills and guidelines will actively mislead agents.
• You need an SLA, LTS guarantees, or a security review process a single-maintainer repository cannot supply.
• You have no AI-agent workflow; the guidelines are still readable docs, but the MCP server and skills are dead weight, and the console/tooling additions may not justify the upgrade.
• The PHP ^8.4 + illuminate ^10.x combination conflicts with your runtime policy — that pairing is the tightest constraint in the manifest and the likeliest source of future friction.

How to Pitch It (Stakeholders)

To executives: Our platform (Tchiowa) is a Laravel-derived framework that standard AI coding tools don't understand, so every AI-assisted change carries rework risk and every new engineer burns time learning conventions that exist only in people's heads. v1.1.0 packages those conventions — 7 guideline documents, 3 task-specific skills, a machine-readable interface that lets AI agents inspect our project directly, plus enforced formatting and static-analysis configs — into one maintained dependency. The business outcome is faster ramp-up, more consistent code review, and AI assistance we can actually trust on our stack. Cost is low: no license, one dependency, an audit pass and a regression test before rollout.

To engineering: This release makes the agent repo-aware. Three things matter: (1) .ai/guidelines and .ai/skills give agents our actual idioms — strict PHP, illuminate/* without the framework, BladeOne, Phinx — instead of generic Laravel patterns that don't compile here; (2) the Node stdio MCP server in mcp/tchiowa/ (built dist/ included, 4 tools) is registered in .mcp.json, opencode.json, and tchiowa.json, so agents can query the project rather than guess; (3) pint.json (Laravel preset) and phpstan.neon at level 5 give us one formatting and analysis baseline. Also in scope: a new DatabaseExporter, 17 console commands, BladeFactory defaulting to MODE_AUTO, and guarded helpers. Two asks before adoption — review the squashed diff (618151d) since it lands a lot at once, and regression-test view resolution and helper call sites, because MODE_AUTO is a behavioral change hiding in a tooling release. Pin the version, keep it out of production runtime, and treat the MCP server as dev-environment only.